home / mcp / autots mcp server
Provides an AutoTS MCP server to run forecasting workflows, manage models, and generate probabilistic forecasts from an MCP client.
Configuration
View docs{
"mcpServers": {
"winedarksea-autots": {
"url": "https://example-mcp.autots.io/mcp",
"headers": {
"YOUR_API_KEY": "YOUR_API_KEY"
}
}
}
}You can run an MCP (Machine-Client Protocol) server for AutoTS to manage forecasting workflows, enabling automated model selection, forecasting, and result exploration from an MCP client. This server config lets you start and control AutoTS forecasting pipelines remotely, integrating with your existing tooling and automation pipelines.
You interact with the AutoTS MCP server by starting the MCP-enabled AutoTS process on your machine or in your environment, then connecting to it from your MCP client. Use the MCP to initiate model templates, load data, run training and evaluation, generate forecasts, and visualize results. The server exposes the forecasting workflow through standard MCP endpoints, allowing you to run large-scale batches across many time series with configurable frequency, validation schemes, and ensemble options.
Prerequisites include a Python environment and a MCP-enabled AutoTS package. Install the MCP-enabled AutoTS package with the following command.
pip install autots[mcp]You can rely on the MCP server for AutoTS to support: remote HTTP-based MCP clients and local STDIO-based clients that launch the AutoTS MCP process. The process includes an installation note and a ready-to-run MCP configuration snippet that demonstrates how to expose the server as an MCP endpoint.
Example MCP configuration snippet (STDIO) to start the server locally:
{
"mcpServers": {
"autots": {
"command": "autots-mcp"
}
}
}Accepts data in long or wide pandas DataFrame formats and trains forecasting models per series, enabling subsequent prediction.
Generates forecasts from fitted AutoTS models, returning point forecasts and optional prediction intervals.
Provides future time-series forecasts across configured horizon lengths for all input series.
Returns performance results and analytics from model validation and cross-validation runs.
Allows saving and loading model templates, enabling repeatable forecasting pipelines.
Visualizes forecast results and model performance alongside historical data.